Meta's year of bold 'superintelligence' bets unlikely to pump profits

By Jaspreet Singh and Aditya Soni

(Reuters) -It's crunch time for Mark Zuckerberg as he pulls out all the stops to stay relevant in Silicon Valley's intensifying advanced artificial intelligence race.

The Meta CEO has sparked a billion-dollar talent war, aggressively poaching researchers from rivals including OpenAI. But as Meta's spending rises, so does the pressure it faces to deliver returns.

For the second quarter, though, Wall Street is bracing for disappointment as the company is set to report its slowest profit growth in two years on Wednesday, rising by 11.5% to $15.01 billion, as operating costs jump nearly 9%.

Revenue, too, likely grew at its slowest pace in seven quarters in that period, up an expected 14.7% to $44.80 billion, according to an average analyst estimate from LSEG.

While Zuckerberg is no stranger to high-stakes pursuits - Meta's augmented-reality unit has burned more than $60 billion since 2020 - his latest push comes with added urgency because of the underwhelming performance of the company's large language Llama 4 model.

He recently pledged hundreds of billions of dollars to build massive AI data centers and shelled out $14.3 billion for a stake in startup Scale AI, poaching its 28-year-old billionaire CEO Alexandr Wang, even as Meta continued to lay off people.

Investors have largely backed Zuckerberg's frenzied pursuit of superintelligence - a hypothetical concept where AI surpasses human intelligence in every possible way - pushing the company's stock up more than a fifth so far this year. 

But they will watch if Meta further increases its capital expenditure for the year after boosting it in April. Alphabet also upped the ante last week, increasing its annual capex forecast by 13% to $85 billion due to surging demand for its AI-powered Google Cloud services.

"We view rising capex as positive given... Meta can become a one-stop shop for many marketing departments," said Ben Barringer, head of technology research at Quilter Cheviot, which holds Meta shares. 

Lagging efforts from Alphabet's Google DeepMind and OpenAI, Meta launched a Superintelligence Lab last month that will work in parallel to Meta AI, the company's established AI research division, led by deep learning pioneer, Yann LeCun. 

To differentiate its efforts, Zuckerberg has promised to release Meta's AI work as open source and touted that superintelligence can become a mainstream consumer product through devices like Ray-Ban Meta smartglasses, rather than a purely enterprise-focused technology.

The strategy plays to Meta's strengths, analysts say, pointing to its more than 3-billion strong social media user base and engagement gains in recent years, driven by AI-enhanced content targeting.

Still, Meta's mainstay advertising market is under threat from advertisers pulling back spending in the face of President Donald Trump's tariffs, and tough competition from Chinese-owned TikTok, whose U.S. ban now seems unlikely.  

Some advertisers may have leaned on proven platforms such as Meta amid the uncertainty, but that will not shield the company from questions over its superintelligence ambitions and how they fit into its broader business strategy, said Minda Smiley, senior analyst at eMarketer.

"While Meta has seen massive gains from incorporating AI into its ad platform and algorithms, its attempts to directly compete with the likes of OpenAI are proving to be more challenging while costing it billions of dollars."

Questions remain about when superintelligence can be achieved, a timeline Zuckerberg admits is uncertain. Meta's LeCun is also a known skeptic of the large language model path to superintelligence.

"Meta's AI strategy today is more cohesive than in 2023, but there's still a sense the company is still searching for direction," MoffettNathanson analysts said.

(Reporting by Jaspreet Singh and Aditya Soni in Bengaluru; Editing by Sayantani Ghosh and Pooja Desai)